mwparserfromhell: KeyError: "000nbsp [while running 'train/Clean content']" while cleaning Arabic data from 20/09/2022
#4
by
uj
- opened
I think i already encountered this error on a previously available dump, but it seems that mwparserfromhell fails on this language only.
While it might be an error from mwparserfromhell, I'd be happy to know where to look at if any of you have ideas.
To reproduce (hopefully), run:
ds = load_dataset("wikipedia", language="ar," date="20220920", beam_runner="DirectRunner")
We had a similar reported issue:
This issue was reported to mwparserfromhell:
And normally it was fixed in the mwparserfromhell library and will be accessible in their next release version 0.7.
See comment: https://github.com/huggingface/datasets/issues/577#issuecomment-1274522440
albertvillanova
changed discussion status to
closed